|
Abstract:
|
We discuss the prerequisites for the design of observational studies to infer the causal effect of a so-called "`immutable characteristic," such as race or gender. Despite their ubiquity, such studies have struggles for theoretical coherence, given (i) the unavailability of a manipulation analogous to a "treatment" in a randomized experiment, and (ii) the danger of post-treatment bias stemming from the fact that immutable traits are determined at conception. We demonstrate that a shift in focus from actual traits to perceptions of immutable characteristics can address both of these issues while facilitating articulation of other critical concepts, such as the timing of treatment assignment and the definition of "race." In a longer version available upon request, we illustrate concepts by discussing the designs of studies of the role of race in trial court death penalty administration.
|
